Note: there is a bug on civit not showing downloads, trust me its getting downloaded.
UPDATE TO THE LATEST COMFY!
9B Model
https://huggingface.co/black-forest-labs/FLUX.2-klein-9B/blob/main/flux-2-klein-9b.safetensors
9B text encoder (choose best for your vram)
https://huggingface.co/Comfy-Org/flux2-klein-9B/tree/main/split_files/text_encoders
vae (same as 4b)
https://huggingface.co/Comfy-Org/flux2-klein-9B/tree/main/split_files/vae
4B Model
https://huggingface.co/black-forest-labs/FLUX.2-klein-4B/blob/main/flux-2-klein-4b.safetensors
4b text encoder (choose best for your vram)
https://huggingface.co/Comfy-Org/flux2-klein-4B/tree/main/split_files/text_encoders
vae (same as 9b)
https://huggingface.co/Comfy-Org/flux2-klein-4B/tree/main/split_files/vae
9B requires qwen 3.8 encoder
4B requires qwen 3.4 encoder
do not confuse this.
i do not believe the model is NSFW after testing.
Description
4b and 9b klein models with condensed workflow
FAQ
Comments (10)
I this error from this workflow, same as the one I got from ComfyUI workflow email, same error. "Got [32, 32, 32, 32] but expected positional dim 64" Any thoughts?
I'm surprised that updating ComfyUI to latest didn't fix this, maybe you're running a different version?
Yea something is blocking your branch pull during your update id assume. Are you running any custom nodes you downloaded outside of comfy? That can block it sometimes (dont ask idk it just does sometimes lol)
And also there is a way around this to update if it is the case, but you have to open a cmd prompt and do some git pulling commands to force update yourself. An LLM will guide you through it easily.
Its definately the update though. Try updating comfy and running again
Im curious, did you put your model in your unet folder or the diffusion model folder?
It goes in diffusion models folder if you put it in unet, that may be the issue. I could be wrong
I was running into this as well - you have to update comfyui to make it work
the same problem , fix it?
@scone317382 updating comfy to the latest will fix it
Updating ComfyUi solved, nice
Can you add inpaint feature with stich image for this workflow for the newbies ?
Both text encoders: qwen_3_8b_fp4mixed.safetensors and fp8 seems not be working, only the full qwen_3_8b model. Can u help me?
















